The Journal of the International Phonetic Association (Vol. 37, N.2) published an article by PhD Rachid Ridouane discussing gemination. It has drawn my attention particularly for the good description of the contrast singleton / geminate not only in word-medial context but also in initial and word-final positions. The study is made for Tashlhiyt Berber ( Korean and Swiss German, particularly the dialect of Bern, appear in the bibliography as other languages for understanding the nature of geminates).

My approach to geminates comes from lenition in Western Romance languages. In comparative terms with Latin, the distinction geminate / singleton is lost, thus producing a readjustment of the plosives: voiceless are voiced and former voiced plosives are lost. The fact that the series of phonemes readjusted follow soft mutation in Celtic languages has been used as a statement to explain lenition in terms of substract. Hence sound change would be not accidental, even less random, but due to language contact.

I must point out something else: in Portuguese lenition reaches the series of nasals and approximants, much as does in Celtic languages. This is a point that I have never found when lenition is discussed.


The main argument against the Celtic explanation would be that lenition in Romance happens within intervocalic context (word-medial) while in Celtic languages mutation is word-initial. To make things more difficult mutation is grammatical in Celtic: it marks gender, for instance.


After reading the article mentioned supra on the Journal of the IPA I consider north-western areas of Africa (with geminates not only intervocalic but also word-initial and final) as I had already noticed for Swiss dialects (I have detected lenition in nasal and approximant series in the dialect of Bern although I could not systematize it) as zones where comparative studies with Celtic could be useful for the better understanding of lenition.

Again, although I do admit it looks like a harder and more difficult way to the finding of universals, I do keep language contact as the best way to explain sound-change.